Old Dominion University October 2012

Transcription

1 Energy Content (Normalized) Old Dominion University October f opt ( b 1) fo Io( R / Q) Qo where b cos V 2 Phase Controller Amplitude Controller Klystron 1.0 Limiter 0.9 CEBAF 6 GeV CEBAF Upgrade Loop Phase Amplitude Detector Amplitude Set Point Cavity , Detuning (Hz) Reference Phase Set Point Phase Detector

2 RF Systems What are you controlling? Cavity Equations Control Systems Cavity Models Algorithms Hardware Generator Driven Resonator (GDR) Self Excited Loop (SEL) Receiver ADC/Jitter Transmitter Digital Signal Processing Cavity Tuning & Resonance Control Stepper Motor Piezo

3 RF Systems are broken down into two parts. The high power section consisting of the power amplifier and the high power transmission line (waveguide or coax) The Low power (level) section (LLRF) consisting of the field and resonance control components RF Control Electronics Cavity Power Amplifier Waveguide/ Coax

4 Think your grandfathers Hi-Fi stereo or your guitar amp. Intensity modulation of DC beam by control grid Efficiency ~ 50-70% (dependent on operation mode) Gain db Frequency dc 500 MHz Power to 1 MW Triode Tetrode

5 Velocity modulation with input Cavity Drift space and several cavities to achieve bunching It is highly efficient DC to RF Conversion (50%+) High gain >50 db CW klystrons typically have a modulating anode for - Gain control - RF drive power in saturation Power: CW 1 MW, Pulsed 5 MW Frequencies 300 MHz to 10 GHz+ A. Nassiri CW SCRF workshop 2012

6 Intensity modulation of DC beam by control grid It is highly efficient DC to RF Conversion approaching 70% Unfortunately low gain 22 db (max) Power: CW to 80 kw Frequencies 300 MHz to 1.5 GHz As a tetrode As a klystron A. Nassiri CW SCRF workshop 2012

7 RF Transistor Efficiencies 40%+ (depends on operating mode) Gain ~ 15 db Power 2 kw at 100 MHz 1kW at 1000 MHz Frequency to 3 GHz Need to combine many transistors to get to very high power 5kW + RF Transistor RF Transistor mounted A. Nassiri CW SCRF workshop 2012

8 HVAC i.e. your house s heating and A/C Set temperature System applies heat/cooling to keep house at the set point Phase/Amplitude Temperature Set Point Set Point Thermostat Proportional Controller HVAC Cavity RF control system Set phase and amplitude System applies phase/amplitude to keep cavity House Cavity field at the set points Sensor/Detector Thermometer Environment

9 Required Field Control to meet accelerator performance: Proton/ion Accelerators: 0.5 o and 0.5% Nuclear Physics Accelerators: 0.1 o and 0.05% Light Source: 0.01 o and 0.01% Loaded Q Optimized for beam loading: Nuclear Physics < 1 ma, ERLs close to zero net current Light Sources 10 s of μa to 100 ma (high current in injectors) Microphonics & Lorentz Detuning: Determined by cavity/cryomodule design and background environment. Master Oscillator/Timing/Synchronization: Determined by application (light sources < 100 fs, frequently < 20 fs). Accelerator Specific: Operational, Reliability/Maintainability Access etc.

10 Amplitude For relativistic particles you want the electron bunch on crest.* If the stability of the cavity field effects the longitudinal emittance.energy spread Field control specification folds back from the end users energy spread requirements Cavity Phase and Amplitude Amplitude error Phase error Phase Degrees * There are special circumstances where the cavities are operated off crest. Examples particle bunching in injectors and FELS

11 Cavity: what's the frequency and Q ext Pulsed, CW, both Cavity Gradient Lorentz detuning Field regulation Residual microphonic background Beam current Klystron/IOT/amplifier effects Other cavity pass bands He pressure drifts Fault Recovery SNS Cavity Mechanical Modes Phase Noise Spectrum Don t Over Build the Control System.KISS

12 Center Frequency Shift (RMS Hz) Determines the feedback gain needed for operation Determines the Q L and the klystron power for lightly loaded cavities C100 HTB Cavity Vibration Frequency (Hz) References: 3

13 Normalized Probablity Microphonic Detuning* RMS Amplitude (Hz) Renascence (stiffened) C s(Hz) ~ Peak Background Microphonics Histogram C100 Renascen ce Frequency Hz *Data Taken in Same Environment Microphonic Impact on cavity power operating at 20 MV/m (100 ma of beam) C100 = 5.3 kw REN = 3.3 kw Potential for cost reduction Utility Power amplifier Cryomodule design should incorporate features to reduce microphonics. References: 3

14 Energy Content (normalized) RF power produces radiation pressures : P = (m 0 H 2 e 0 E 2 )/ Pressure deformations produce a frequency shift : f = K L E 2 acc CEBAF 6 GeV CEBAF Upgrade Outward pressure at the equator Inward pressure at the iris , Detuning (Hz) The Quadratic relationship with Gradient becomes an issue at the high gradients (15+ MV/m) needed for new accelerators References: 4,5

15 The simplest representation of a cavity is a parallel LRC circuit V (t)=v e Coupling to the cavity can be represented by a transformer and then reduced to the following circuit Z=( + +jωc) R jωl c c jωt L R C I G (t) Z O L R C I G (t) Z G L R C 1:k where References: 1 Z G 2 k ZO And the coupling is defined 1 k 2 R Z O

16 R sh r /Q Shunt Impedance Ω R L Loaded Shunt impedance Geometry Factor Ω Intrinsic Quality Factor Q o Ω E l Electric Field V/m Loaded Quality Factor Electrical Length m Generator Power W Q L P g f o Cavity Frequency Hz P c Cavity Dissipated Power W W f f o b I b Stored Energy J Beam Phase Coupling Coefficient Beam Current A Tuning Angle V c Cavity Voltage V f o Cavity Detuning Hz Cavity Bandwidth Hz f Static Detuning Hz Microphonic Detuning Hz m

17 Substitutions for L & C o 1 LC C Q O R o Shunt Impedance Loaded Q and Shunt Impedance R sh V P Q 2 2 c c or from circuit theory Rsh C 2PC L QO 1 R L V ( r / Q) Q L Generator Power P Vc (1 ) IOR L IOR L 1 cos tan sin RL 4 Vc Vc g b b Detuning Angle Total Detuning References: 1, 27 f tan 2Q L 2Q L f f f f o m O O

18 I (t)=i e g g jωt L R C I (t)=i e b b jωt V (t)=v e c c jωt Beam and RF generator are represented by a current source I b produces V b with phase which is the detuning angle I g produces V g with phase Cavity Voltage V c = V g V b R R L L V C= IGcos IBcos References: 1,

19 R R L L V C= IGcos IBcos 2 2 V C : V B V G I G I B Im Re V B R L 2 I B cos P G 1 R I 16 2 Sh G V 2 (1 ) 1/2 1/2 G ( P ) cos 1/2 GRL References: 1, 2, 19, 20

20 From generator power We can determine the Minimum power needed Substituting in for b where Some assumptions Differentiating Pg with respect to Q L and setting it to zero leads to optimum Coupling Q V f P bq Q bq 2 c 2 g 1 L cos b 2 L L sin b r 4 Q fo L Q Q IO r b V Q O c Q L Q therefore Q f 4( ) f 4( ) sin 2 2 QLopt b b fo fo O L b 1/2 2 For beam on crest b =0 And if I o = 0 References: 1, 20 Q Lopt f O 2f Q Lopt IO r f ( ) 4( ) Vc Q fo 2 2 1/2 And if I o >> 0 Q V / I ( r / Q) Lopt O

21 Beam on Crest Q f /2 f Lopt o Cavity Detuning 0 Hz Power Hz 10 Hz I = 10 ma Hz E E E+08 References: 1 Q L

22 Beam on Crest Power Q V / I ( r / Q) Lopt Cavity Detuning 0 Hz 5 Hz 10 Hz 20 Hz E E E+08 O I = 1 ma References: 1 Q L

23 f f f o m f o is the static detuning or slow detuning..< 1 Hz Use active mechanical tuning to control cavity frequency f m is the fast detuning do to microphonics >10 Hz Use active electronic feedback for field control Gray area between 1 and 10 Hz

24 Classic Plant-Controller can be used to model the RF control system Cavity Can be modeled a variety of ways. Lorentz force can be added as a non-linear element Mechanical modes can also be included Power Amplifier (Klystron, IOT, Solid State, etc.) Saturation effects Hardware Group delay (line delay, processing latency etc.) Modeling software such as Matlab/Simulink has made this rather easy.

25 X(s) E(s) Controller K(s) Plant H(s) Y(s) G(s) Need to solve for close loop transfer function: Y/X E( s) X( s) G( s) Y( s) Y( s) K( s) H( s) E( s) Sensor Y ( s ) H ( s ) K ( s ) X( s) 1 H( s) K( s) G( s) References: 28

26 Want Output/Input (V/I) assume Parallel RLC circuit with current source V V I jcv R jl V Convert to s domain V( s) Sub for L&C References: 28 jrl 2 I RLC jl R Gs ( ) srl 2 I( s) s RLC sl R Gs ( ) s QL s ORL s 2 O 2 O QL I L R C I(s) V(s) G(s) V( s) G( s) I( s) V

27

28 We are only interested in the time scale of the detuning microphonics. Therefore we can model the cavity at baseband using a complex envelop by separating V(t) and I(t) into I and Q terms. j t ( ) ( ) ( ) V( t) V ( t) jv ( t) e R From the 2 nd ODE for a driven oscillator Want the state form I x Ax Bu j t I t I t ji t e V t V t V t R I t O 2 O ( ) ( ) ( ) L ( ) QL QL where A and B are state Matrixes Ignoring 2 nd derivative terms (small compared to lower order terms), assume that / o ~ 1 and = o - is the cavity detuning O 2Q L O A VR O VR VR V I RL I O x R 2QL 2Q VI L 2Q L O O V V V R I 2Q 2Q I I R L I L L References: 18, 28, 35 R I R B L 2 Q 0 O L R L 0 O 2 Q L I u II R

29 Quickest way to build an oscillator is to design a control system...various authors An unstable condition occurs when Gain >1 and the phase shift through the system is more than 180 degrees. Know your poles! Since we are working with sc cavities the first pole will be the ½ cavity bandwidth ( Hz) Each pole contributes 90 degrees starting a decade before the pole and ending a decade after. Need to do a phase delay budget all the way around the system, RF Controls-Klystron-waveguide-cavitycables etc.

30 There are many methods (Routh and Nyquist) to determine stability of a control system. Matlab will actually perform this analysis. Amplitude Bode Diagrams Graphically intuitive Give Phase and Gain margin Instabilities start as small peaks Phase

31 Reference X(s) E(s) Error Signal PID Controller K(s) System Delay Cavity G(s) H(s) PID Function 2 KI KDs KPs KI K( s) KP KDs s s System Delay (Four Poles at 100 khz) Hs ( ) ( ) 5 s Cavity Bandwidth ~ 25 Hz Gs ( ) 157 s 157 References: 28

32 Cavity pole Half power (-3dB) In G(s) Out G(s) = a/(s+a) 25 Hz Gs ( ) 157 s 157

33 Proportional Term P K e() t K P Kp ~ proportional gain e ~ error Integral Term P K I KI e( ) d s t 0 K I ~ integral gain Derivative Term de() t D KD KDs dt KD ~ derivative gain I In K(s) Out 2 KI KDs KPs KI K( s) KP KDs s s References: 28

34 K P = 100 K I = 0 K D = 0 Gain = 1 Gain Margin ~ 25 db Phase Margin ~ 90 o Phase = 180

35 K P = 100 K I = 10 K D = 0 Effect of Integrator

36 If the required closed loop control error signal is << than the open loop error signal (with out proportional gain), then the proportional gain (K p ) can be approximated by the following K p = (Open Loop error signal)/(required closed loop error signal) This is a handy rule of thumb

37 Numerical based model, coded in C Incorporates feedback and feed forward Cavity modeled at baseband as a low pass filter References:

38 Matlab/Simulink based Cavity represented in state space A library of model blocks is available Feedback/Feedforward Lorentz Force GDR and SEL Various other algorithms (Kalman, Smith Predictor) References: 32

39 Cavity representation is simplified to quadrature components using low pass filter (cavity bandwidth/2). Lorentz Force detuning, microphonics and tuners function are incorporated as a frequency modulators. Baseband simulation, means sampling time for processing can be large (1usec) thus simulation speed is high. Rotation matrix for quadrature components to reflect detuning frequency Microphonics: External noise generator References: 29

40 Generator Driven Resonator (GDR) Vector Sum (Flash.ILC) Feed forward (for pulsed systems) Self Excited Loop (SEL) Microphonics (detuning) compensator Other Control Algorithms Kalman Filter Adaptive Control (LMS)

41 Essentially an extension of the classic Controller Plant model Easily adaptable to I and Q domain for digital control. Advantages - Where fast/deterministic lock up times are critical i.e pulsed systems. Disadvantages Not frequency agile needs tuning elements to keep cavity close to reference High Q machines with high microphonic content and large Lorentz detuning could go unstable Phase Set Point Reference Phase Controller Amplitude Controller SC Cavity Klystron Amplitude Set Point References: 30

42 Cavity acts as a tank circuit for the feedback. Much like a VCO in a PLL. By adjusting loop phase, loop is forced to operate at the reference frequency. No free lunch cavity still must be locked close to the reference to avoid saturating the power amplifier Reference Limiter Loop Phase Phase Controller Amplitude Controller Amplitude Set Point SC Cavity Klystron Advantages - High Q L Cavities - Systems with large Lorentz detuning Disadvantages - Slow lock up time Phase Set Point Phase Detector References: 30

43 When locked the methods are equivalent. GDR only operates in the DC domain i.e. the converted vector does not spin only has angle and magnitude dependence. A digital SEL must be able to handle a spinning vector when the system is not locked i.e. there is an t term that must be accounted for when the SEL is in oscillation mode. DSEL capture range only limited by digital filter (~ 100 khz) Easy configuration switch between GDR and SEL Ultimately some hybrid digital SEL/GDR of the two may be the solution.

44 The standard method for RF control up until ~ 1995 Still have relevance in accelerators that need little adjustment Advantages Economical Simple design Disadvantages flexibility Pulsed beam loading RF/analog signal processing parts harder to find Recent Installations Daresbury, ELBE/Rossendorf

45 Overwhelmingly the majority of new RF controls employ digital feedback Advantages Flexibility Flexibility did I mention flexibility?? SNS RF Controls Disadvantages Complexity this is relative Installations too numerous to list Cornel RF Controls

46

47 Digital Board PC-104 FPGA Fast ADCs Receivers 56 MHz Clock RF Board RF = 1497 MHz, IF = 70 MHz 5 Receiver Channels One transmitter Digital Board FPGA: Altera Cyclone 2 EP2C35 16 bit ADC s & DAC Quadrature sampling at 56 MHz Ultra low noise 56 MHz clock IOC PC104/ RTEMS/EPICS RF Board Transmitter

48 Heterodyne scheme Noise Figure: Typically large cavity signal so not an issue Signal to Noise Ratio (S/R, SNR): Ultimately determines field control. Dominated by ADC Component linearity effects dynamic range and signal accuracy Clock and LO Phase Noise/Jitter can impact control RF Mixer IF Amplifier ADC SC Cavity LO Band Pass Filter Anti-Alias Clock

49 Errors in the receiver chain are driven into the cavity by feedback gain in the controller Quantization noise in the ADCs Harmonic distortion aliased back into the passband Quantization error Channel to channel crosstalk Nonlinearities in the down converter You can t recover from a bad receiver design! B. Chase SRF

50 Drawing courtesy of Marki Microwave y IF (t) y RF (t) y LO (t) y RF (t) y IF (t) y LO (t) Down conversion conserves phase, thus lowering jitter sensitivity by: B. Chase SRF f IF f RF

51 Mixers are based on diodes which are the nonlinear devices used for current switching This process generates unwanted harmonics Diodes are not perfect switches and distort the IF signal producing harmonics Ima ge spur Choice of LO determines filter requirement B. Chase SRF 2011

52 Should be considered for frequencies < 500 MHz Clock jitter/phase noise is crucial ADC must have small aperture jitter (< 300 fs) Benefits Economy. $$$ Simpler receiver for multiple frequencies Simpler Master Oscillator (MO) distribution RF Amplifier ADC SC Cavity Band Pass Filter Anti-Alias Clock

53 Loop controller processing is usually done at baseband as these signals are needed for diagnostics Most modern SCRF LLRF systems digitize RF or IF signals and frequency translates them to baseband with Digital Down Converters (DDC) built with Numeric Controlled Oscillators Other options are direct conversion or direct IQ conversion M f s F if N IF M/N is rational, the NCO table can be small If M/N = 4 then the phase step is 90 degrees ( IQ sampling) B. Chase SRF 2011

54 Amplitude Amplitude Sample a frequency at four points Can either harmonic sample or sub harmonic Example: A signal at 50 MHz. Harmonic Sampling frequency is 200 MHz Sub harmonic Sampling or 40 MHz (200/5) Easier to sample at lower frequencies Phase Phase

55 Factors in heavily in determining your systems S/N and dynamic range Calculate how many bits that you need then add two! Additional Cost will be minimal S/N determined by, number of bits, quantization error, sample rate and system jitter. Parallel Architecture clock speed > 50 MHz Advantage: Wide bandwidths > 700 MHz Serial Architecture clock speeds < 50 MHz Advantage: multiple channels, less pins needed, smaller FPGA

56 Whether you are down converting to an IF or Direct sampling the systems S/N needs to be determined to meet field control specifications. The S/N of an ideal ADC is given by the following equation S / N [6.02N 1.76] Number of bits S/N (ideal) db S/N (real) db S/N determines the lower limits of the field control you can expect.a light source would need more bits

57 References: 26 When we factor in linearity and clock jitter the S/N decreases even more. In the case of clock jitter, the conversion error can be seen below

58 S/N Real world S/N equation e 2 2V NOISErms S / N 20log10 ( ot jrms ) N N Where o is the analog input frequency (2pf), t jrms is the combined jitter of the ADC and clock, 70 e is the average differential nonlinearity (DNL) of the ADC in LSBs, Bit ADC S/N vs Input Frequency 2 V NOISErms is the effective input noise of the ADC in LSBs and N is the number of ADC bits fs 200 fs 300 fs References: E E E+09 Frequency (Hz)

59 Phase Noise (dbc/hz) Phase Noise is the parameter used in the communication industry to define an oscillators spectral purity. Timing Jitter is the parameter most used by accelerator designers in describing beam based specifications and phenomena. Phase noise spectrum can be converted to a timing jitter using the formula A/ p f O Where A is the area under the curve References: 22, 23, Signal Source Phase Noise Agilent 850 MHz Wenzel 70 MHz RS 1GHz ,000 10, ,000 1,000,000 Frequency (Hz) Signal Source Table: Signal Source Jitter Integrated Phase Noise (dbc/mhz ) Jitter (RMS) Agilent(1497 MHz) fs Rhode Schwarz(1497 MHz) fs Wenzel (70 MHz) fs

60 Parameter Specification Value Imposing Quantities S/N 72 db 0.1 degree resolution, 0.01% gradient accuracy Bandwidth 8 MHz Latency, S/N, temperature stability Latency 100 ns Control BW Noise Figure (NF) 52 db, BW = 100 khz S/N for phase resolution Linearity 0.01% F.S. Stability, accuracy Dynamic Range +54 dbm IIP3 Gradient range Channel Isolation 67 db Phase, gradient resolution/accuracy In-band intermodulation distortion (IMD) 67 dbc THD

61 Amplitude Error Receiver S/N determines minimum residual amplitude control Amplifiers Mixer ADC 3.0E E E E-04 Measured Amplitude Error vs. Proportional Gain Integral Gain I = 0 I = 1 I = 2 Linear components needed for stability and accuracy over large dynamic range It is possible to improve S/N, through process gain, but at the expense of control bandwidth and ultimately stability (latency). References: E E E E-05 Receiver Floor RF=1497 MHz IF = 70 MHz ADC = 14 bits Proportional Gain

62 Highly dependent on the reference (LO/IF) and subsequent board level clock Linear components needed to minimize AM to PM contributions ADC aperture jitter ~ 100 fs Some ADC linearity can be improved with near quadrature sampling Open loop Microphonics Receiver Floor Phase Noise of Open and Closed Loop. Bright Yellow is Closed loop. References: 30

63 FPGA or DSP? For fast processing Field Programmable Gate Array (FPGA) wins DSP a little more flexible but FPGAs not far behind Suggest FPGA followed by a small CPU (ColdFire or PC104) of some sort (best of both worlds!) For FPGA calculate your gate needs then double or triple the size to be safe. Cost is a wash. FPGA: Xlinx Altera. Both are used by the accelerator community DSP: Texas Instruments Analog Devices dominate industry

64 I /Q easiest to implement Non I/Q sampling eliminates ADC non-linearities Frontend processing can be decimation followed by FIR.not efficient Better method is to use a Cascaded-Integrated- Comb (CIC). N-stage cascaded integrator-comb (CIC) filter (decimator) PID algorithm simplest for CW applications PID can be enhanced with feed-forward or gain scheduling for use in pulsed systems.

65 IIR Filter: Best for realizing analog filters Need to decimate to Fs/Fo of ~ 100 to insure stability Can have high-gain and round off errors Amplitude Phase Oversampling can improve S/N by ~20log(N 1/2 ), but at the expense of control bandwidth Example: clock = 50 MHz and bandwidth needed is 1 MHz. The potential S/N improvement would be 17 db Cavity Emulator using IIR Filter References: 30

66 IF direct I&Q sampling Digital filtering PID controller for I and Q values Rotation matrix Single DAC generating IF signal References: 31

67 IF direct I&Q sampling digital filtering I&Q to Phase&Magnitude - COordinate Rotation DIgital Computer (CORDIC) SEL mode Microphonics Compensation single DAC generating IF signal References: 31

68 COordinate Rotation DIgital Computer Iterative method for determining magnitude and phase angle Avoids multiplication and division N bits +1 clock cycles per sample Can also be used for vectoring and linear functions (eg. y = mx + b) Exploits the similarity between 45 o, 22.5 o, o, etc. and Arctan of 0.5, 0.25, 0.125, etc. Multiplies are reduced to shift-andadd operations Angle Tan ( ) Nearest 2 -N Atan ( ) cos sin sin x', y' x, y References: 33, 34 cos x y i1 i1 K K i i x i y i yi x i d i d i 2 2 i i

69 DAC typically the same number of bits as the ADC DAC needs to be fast enough to support IF generation or RF Conversion if Direct Sampling Mixer specification (notably IP3) can be relaxed from cavity mixer. Depending on pre-amp (solid state) and power amplifier (Klystron/IOT), you may need to filter out LO. DAC Quadrature Modulator I IF Klystron /IOT DAC Q Band Pass Filter LO RF SC Cavity DAC IF RF Band Pass Filter LO Klystron /IOT SC Cavity

70 Single DAC can eliminate Quadrature Modulator See Larry Doolittle web page: T=N t t N(f o+1) f 1 =Nfo Concept use one of the harmonics out of your ADC for your IF frequency. For a 10-X system two disadvantages to using second or third harmonic frequencies are: Small signal content. Analog filter requirements. f f

71 T=4 3-X DDS t t 8 MHz Filter 5f o f =4f o= 56 MHz 70 MHz Ratios of f3 to f1 is 1:5. 70 MHz component is 14 MHz away from nearest neighbor. Commercial drop in 8 MHz BW filter available for $30. One can show that the harmonic contains the proper phase signal and is: 2pf B Asin2p kf f t where k 0,1, 2... Asin k S f f

72 Goals of a resonance control system Keep the cavity as close as possible to the reference frequency, ultimately minimizing forward power. Compensates for Lorentz detuning for pulsed systems Minimize microphonics to assist electronic feedback Reliable and maintainable Tuning Methods Stepper Motor: speed < 1 Hz Piezo tuner: speed > 1 Hz References: 19 Prototype tuner for CEBAF Upgrade

73 Example CEBAF (1986) to CEBAF Upgrade (2008) Frequency Gradient Bandwidth Lorentz detuning Range Resolution Tuning method Drive CEBAF 1497 MHz 5 MV/m 220 Hz 75 Hz +/- 200 khz 10 Hz Ten/Comp Stepper CEBAF Upgrade 1497 MHz 20 MV/m 50 Hz 800 Hz (est.) +/- 200 khz < 1Hz Tension Stepper & PZT Gradients increased five fold in 20 years bringing Lorentz detuning into play and the need to keep klystron size small. Tuner designs responded with faster tuning methods (PZT) and increased resolution. Future/Now: active mechanical compensation of microphonics References: 19

74 Stepper Motor: Recover cavity from large excursions associated with down time activities or Cryogenic trips. Keeps the Fast Tuner centered Control can be slow < 1 sec Stepper motor Stepper Motor Driver JLAB Upgrade Tuner assembly

75 Piezo-Electric Tuner (PZT): Large Industrial base for Piezo and electronics Recover or compensate for Lorentz Detuning (Feed Forward or Feedback) Minimizes small changes in resonance do to He pressure. Speed < 1 ms Control logic embedded in FPGA or fast DSP Warm Stroke greater than Cold Stroke Has been demonstrated to minimize cavity microphonics

76 Scissor jack mechanism Ti-6Al-4V Cold flexures & fulcrum bars Cavity tuned in tension only Attaches on hubs on cavity Warm transmission Stepper motor, harmonic drive, piezo and ball screw mounted on top of CM Openings required in shielding and vacuum tank No bellows between cavities Need to accommodate thermal contraction of cavity string Pre-load and offset each tuner while warm Evolution of the tuner! References: 19

77 Stepper Motor 200 step/rev 300 RPM Harmonic Drive Gear Reduction = 80:1 Low voltage piezo 150 V 50 mm stroke Ball screw Lead = 4 mm Pitch = mm Bellows/slides axial thermal contraction References: 19

78 Resolution/Deadband < 2 Hz Drift due to Helium pressure fluctuations References: 3

79 Frequency Difference (Hz) Piezo tuner voltage vs frequency Difference From Maximum for FEL03-6 at 10 MV/m First cycle Second cycle Voltage (V) References: 3

80 Mechanism Stainless steel rocker arm and drive rod Attaches to chocks on cavity Attaches via flexures and threaded studs to helium vessel head Cavity tuned in compression or tension Cold transmission compressive/tensile force on drive rod Stepper motor and piezo external to vacuum tank Bellows on vacuum tank Need to accommodate relative thermal contraction of cavities Allow tuner transmission to float (unlocked) during cooldown Pre-load each tuner while warm, account for vacuum loading on bellows References: 19

81 ff 0 (khz) ff 0 (khz) pulling pushing Displacement (mm) Cycle second third Piezo drive signal (V) References: 19 T. Kandil LINAC 2004

82 2. TESLA Cavities and Auxiliaries as ILC Baseline Design ± 1 mm fine tuning (on cavity) ΔF on all piezo (sum) 3.5 kn 1 khz fast tuning 3 µm cavity displacement 4 µm piezo displacement 4 µm piezo displacement Δ F on all piezo 11.0 N ~1 Hz resolution (sufficient if <5Hz) Stiffeners bars could be used in working cond. as safety devices. Piezo Ti or SS ring welded on the tank Leverage arm References: 19, 21

83 Mechanism All cold, in vacuum components Stainless Steel frame Low voltage 40 mm Noliac Piezo Attaches to helium vessel shell Phytron stepper motor with planetary gear box Cavity tuned in tension or compression blades provide axial deflection References: 21

84 Piezo tuner is basically a capacitive device. Current is only needed during dynamic tuning dv I C jcv dt Which ultimately determines the size of your power supply. Example: JLAB PZT Voltage Range : 0 to 150 volts (full stroke) Capacitance: 21 uf (warm) Tuning range: 0 to 2000 Hz (0 to 150 Vdc) We have observed a mechanical perturbation at a frequency of 8 Hz The perturbation effect on cavity detuning is 60 Hz (@ 1497 MHz)

85 Goal is to damp the 8 Hz Perturbation Since it has 60 Hz detuning the PZT voltage will need to be compensate over the range of 60 / 2000 V 4.5 Volts Putting this into the previous equation (I=jCV) gives us I 2p 8Hz 21uF ma Which is a relatively small power supply. As a comparison a system with a 1000 volt PZT (same stroke), and a similar disturbance the current required is 31.7 ma Some thought needs to go into PZT voltage for safety too!

86 Cavity Detuning (Hz) PZT Current Needed to control Detuning ma ma 20 ma 10 ma Microphonic Frequency (Hz) Assumes 150 Vdc PZT with 2000 Hz full stroke

87 Normalized Output (S21) Features Commercial APEX piezo driver amplifier. 0 to 150 V at 50 ma. Bandwidth (FS): 10 Hz Packaging: 3U Eurocard x PZT Amplifier Transfer Function APEX Amplifier PZT Drive 36 Vpp 0.1 PZT Drive 5 Vpp PZT Drive 150 Vpp Freq(Hz)

88 Commonly used to adjust coupling Could also be used to compensate for detuning Issues: Part of the waveguide becomes part of the resonant system Speed for dynamic control of microphonics Good for tuning Q L higher not so much lower JLAB uses a three stub tuner to adjust cavity Q ext References: 19

89 Solution for linacs powering multiple cavities from one power source, especially cw designs. Need to compare cost and complexity between RF power source and power supply for ferrites Vector Modulator Phase Shifter References: 19

90 Challenges - Thoughts Design LLRF with respect to what is needed by the accelerator and the cryomodule. Example: A proton/ion LLRF control system doesn't need light source precision! Field control requirements beyond 0.05 o and.01% control are pushing the limits of the receiver hardware. Trade offs between process gain (increased latency) and loop gain need to be made to reach beyond these values. A lot of room to grow in fast mechanical control. Could have big pay off in reduced amplifier power Good Luck!

91 [1] L. Merminga, J. Delayen, On the optimization of Qext under heavy beam loading and in the presence of microphonics, CEBAF-TN [2] M. Liepe, et al, Proceedings of the 2005 Particle Accelerator Conference, Knoxville, USA [3] K. Davis, T. Powers, Microphonics Evaluation for the CEBAF Energy Upgrade, JLAB-TN [4] D. Schulze, Ponderomotive Stability of RF Resonators and Resonator Control Systems, KFK 1493, Karlsruhe (1971); ANL Translation ANL-TRANS-944 (1972). [5] J. R. Delayen, Phase and Amplitude Stabilization of Superconducting Resonators, Ph. D. Thesis, California Institute of Technology, [6] T. Plawski, private conversations, [7] A.S. Hofler et al, Proceedings of the 2004 Linear Accelerator Conference, Lubeck, Germany [8] A. Neumann, et al, Proceedings of the 2004 European Particle Accelerator Conference, Lucerne, Switzerland. [9] C. Hovater, et al, Proceedings of the 2007 Particle Accelerator Conference, Albuquerque NM, USA [10] L. Doolittle, Proceedings of the 2007 Asian Particle Accelerator Conference, Indore, India [11] S. Simrock, et al, Proceedings of the 2006 Linear Accelerator Conference, Knoxville, TN USA [12] J. Musson, private conversations, [13] F. Ludwig et al, Proceedings of the 2006 European Particle Accelerator Conference, Edinburgh Scotland [14] U. Mavric and B. Chase, Microwave Journal, Vol. 51, No. 3, March 2008, page 94 [15] L. Doolittle et al, Proceedings of the 2006 Linear Accelerator Conference, Knoxville, TN, USA [16] M. Liepe, et al, Proceedings of the 2001 Particle Accelerator Conference, Chicago, IL, USA [17] K. Davis, J Delayen, Proceedings of the 2003 Particle Accelerator Conference, Portland USA [18] T. Schilcher, Vector sum control of pulsed accelerated field in Lorentz force detuned superconducting cavities, PhD thesis, Hamburg 1998 [19] J. Delayen, LLRF Control Systems Tuning Systems, 2007 SRF Workshop, Beijing China, October 2007 [20] M. Liepe, Operationa Aspects of SC RF Cavities with Beam, 2007 SRF Workshop, Beijing China, October 2007 [21] A. Bosotti, et al, Full Characterization of Piezo Blade Tuner for Superconducting RF Cavities, Proceedings of EPAC08, Genoa Italy [22] Walt Kester, Converting Oscillator Phase Noise to Time Jitter, Analog Devices MT-008 [23] F.L. Walls et al Tutorial: Fundamental Concepts and Definitions in PM and AM Noise Metrology; Discussion of Error Models for PM and AM Noise Measurements; State-of-the-Art Measurement Techniques for PM and AM Noise (3 parts), pg 71 [24] NIST, Physics Laboratory, Time and Frequency Division, [25] L. Doolittle, Plan for a 50 MHz Analog Output Channel, [26] R. Reeder et al, Analog-to-Digital Converter Clock Optimization: A Test Engineering Perspective, Analog Dialogue Volume 42 Number 1 [27] H. Padamsee, et al, RF Superconductivity for Accelerators, John Wiley & Sons, 1998, pp. 47 [28] K. Dutton, et al, The Art of Control Engineering, Addison Wesley, 1998 [29] T. Plawski et al, RF System Modeling for the CEBAF Energy Upgrade Proceedings of PAC09, Vancouver, May 2009 [30] C. Hovater, RF CONTROL OF HIGH Q L SUPERCONDUCTING CAVITIES Proceedings of LINAC08, Victoria BC [31] T. Allison et al, A Digital Self Excited Loop for Accelerating Cavity Field Control, Proceedings of PAC07, Albuquerque, June 2007 [32] A. Vardanyan, et al, An Analysis Tool for RF Control for Superconducting Cavities,, Proceedings of EPAC02, Paris France [33] Jack E. Volder, The CORDIC Trigonometric Computing Technique, IRE Transactions on Electronic Computers, September 1959 [34] R. Andraka, A Survey of CORDIC Algorithms for FPGA Based Computers, FPGA '98. Proceedings of the 1998 ACM/SIGDA Feb. 1998, Monterey, CA. [35] A. Neumann, Compensating Microphonics in SRF Cavities to Ensure Beam Stability for Future Free Electron Lasers, PhD Thesis, Berlin 1998

92 Y Binary search, linked to sgn(y) d i 1, 1, if if y y i i 0 0 Successively add angles to produce unique angle vector X i1 Resultant lies on X (real) axis with a residual gain of 1.6 z i z i d i arctan(2 ) i d i arctan(2 i )

93 DAC Quadrature Modulator I IF Klystron/IOT DAC Q Band Pass Filter LO RF SC Cavity DAC IF RF Band Pass Filter LO Klystron/IOT SC Cavity

94 2*LO1*RF"image"Spur Mixing products are generated in the mixer and are sourced to all three ports The 2LOx1RF spur is reflected off of the RF bandpass filter and then mixed down on top of the IF Drawing courtesy of Marki Microwave B. Chase SRF